Why can't I achieve 100% accuracy when my testing audios are just from the training data? Isn't 100% accuracy achievable? And why does my database_name.align shows asterisks on some of the decoded audios.
Like
"This is what I am saying"
" * ** * I am saying" <------ It becomes like this.
I have a java program similar to the transcriber demo. I used my audio files from my training folder to use it as input but recognition is very low. i tried recording the same text of the input audio and it surprisingly transcribed it perfectly.
I need the trained audios to be transcribed properly with 100% accuracy by my JAVA Program. This is part 2 of my testing if my speech recognition is working properly.
for answering all my questions sir, thank you very much.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Why can't I achieve 100% accuracy when my testing audios are just from the training data? Isn't 100% accuracy achievable? And why does my database_name.align shows asterisks on some of the decoded audios.
Like
"This is what I am saying"
" * ** * I am saying" <------ It becomes like this.
Why?
Last edit: Leimiaoren 2016-01-07
There are many reasons for bad accuracy. You need to provide the data to reproduce your problem.
Asterisk denote missed words in alignment.
This is my training folder sir nickolay.
I have a java program similar to the transcriber demo. I used my audio files from my training folder to use it as input but recognition is very low. i tried recording the same text of the input audio and it surprisingly transcribed it perfectly.
I need the trained audios to be transcribed properly with 100% accuracy by my JAVA Program. This is part 2 of my testing if my speech recognition is working properly.
for answering all my questions sir, thank you very much.
I finally found whats wrong. it was my data all along. It was resampled to 44khz by my friend.